Improved PET Image Analysis Designed to Optimize Radiotherapy Treatments

A Spanish scientist has implemented 12 algorithms to process medical images generated by positron emission tomography (PET). The new process has promise and could become a useful imaging tool. A new graphics interface has also been designed that will enable use this technique in clinical practice comfortably, quickly, and intuitively.

Elena Prieto-Azkarate, a graduate in telecommunications engineering at the NUP/UPNA-Public University of Navarre (Spain) and a member of the nuclear medicine service of the University College Hospital of Navarre, reported her achievement in her PhD thesis, read at the NUP/UPNA (Public University of Navarra; Spain). The research comes from within the field of biomedical engineering and it was conducted through collaboration between the nuclear medicine service of the University College Hospital of Navarre (Spain) and the NUP/UPNA-Public University of Navarre. The PET technique makes it possible to gather molecular images that provide important data about the biologic and metabolic behavior of tumors. Recent years have seen the emergence of great interest in the potential of images of this type in the planning of courses of radiotherapy treatment. In this planning the most critical process is the exact delimitation of the volume of the tumor requiring treatment.

Dr. Prieto reported that her development has improved the segmentation technique of PET images. “The segmentation of an image,” she explained, “is an image processing technique that allows objects to be delimited— tumors, in this case. This thesis sought to conduct research into and develop new segmentation techniques to turn PET into a reliable alternative for treatment planning in radiotherapeutic oncology.”

Specifically, the researcher has worked on one specific image segmentation technique: thresholding. “Automatic thresholding allows the edges of the tumor to be automatically delimited over the image, which is of tremendous importance since the spatial resolution of images of this kind hampers manual delimitation,” Dr. Prieto noted.

The analyzed images collected from patients were obtained at the University College Hospital of Navarre thanks to funding from the Carlos III Institute of Health (Spanish Ministry of Health and Consumption) through a FIS project, and the company Siemens HealthCare (Erlangen, Germany). Two different PET scans were used in the research to acquire the images so that the automatic thresholding could be assessed under a wide range of conditions. All the images obtained have been made available over the Internet to provide a shared authentication framework for any segmentation technique.

“The results on experimental images have been highly promising,” Dr. Prieto concluded, “and it has been possible to achieve an improvement over the standard technique in the clinical images coming from high-resolution PET tomographs.” Moreover, “the technique used could be highly useful when segmenting images obtained in the state-of-the-art clinical PET/CT tomographs.”

The study’s findings were published in February 2013 issue of the journal Clinical Nuclear Medicine.

Related Links:
NUP/UPNA-Public University of Navarre
Siemens HealthCare